Back to EveryPatent.com
United States Patent |
6,052,462
|
Lu
|
April 18, 2000
|
Double talk detection and echo control circuit
Abstract
In a full duplex audio telecommunication system, it is desirable to
determine whether a transmitted signal contains a signal component
generated exclusively at one end of the communication path. A removal
filter is connected to receive a Far-End signal and to remove a signal
component therefrom. A detection filter is connected to receive a Near-End
signal and to detect the presence of a signal component corresponding to
the component removed from the Far-End signal. A decision stage connected
with the detection filter responds to the detection filter by actuating a
control signal when signal activity in the selected signal component is
detected. The control signal is used to control adaption of an AFIR filter
of an echo canceler to prevent adaption during intervals of Near-End
speech activity. The selected signal component preferably comprises a
predetermined frequency range.
Inventors:
|
Lu; Youhong (Sterling, VA)
|
Assignee:
|
Tellabs Operations, Inc. (Lisle, IL)
|
Appl. No.:
|
890923 |
Filed:
|
July 10, 1997 |
Current U.S. Class: |
379/406.08 |
Intern'l Class: |
H04M 001/58 |
Field of Search: |
379/411,410,406
|
References Cited
U.S. Patent Documents
3935403 | Jan., 1976 | Penicaud et al.
| |
4177361 | Dec., 1979 | Birck.
| |
4670903 | Jun., 1987 | Araseki et al. | 379/411.
|
4747132 | May., 1988 | Ibaraki et al.
| |
4852081 | Jul., 1989 | Bonnet et al.
| |
4891801 | Jan., 1990 | Marcos et al.
| |
4903247 | Feb., 1990 | Van Gerwen et al.
| |
4965822 | Oct., 1990 | Williams.
| |
5014263 | May., 1991 | Vairavan et al.
| |
5151937 | Sep., 1992 | Chujo et al.
| |
5164989 | Nov., 1992 | Brandman et al.
| |
5317596 | May., 1994 | Ho et al. | 375/14.
|
5319636 | Jun., 1994 | Long et al.
| |
5390250 | Feb., 1995 | Janse et al.
| |
5463618 | Oct., 1995 | Furukawa et al.
| |
5535194 | Jul., 1996 | Ashley et al.
| |
Primary Examiner: Isen; Forester W.
Assistant Examiner: Pendleton; Brian
Attorney, Agent or Firm: Dann, Dorfman, Herrell & Skillman
Claims
That which is claimed is:
1. An apparatus for detecting Near-End audio activity in a two-way
telecommunication system connected between a Near-End and a Far-End, the
apparatus comprising:
a first input terminal for receiving a first telecommunication signal from
the Far-End;
a removal filter connected with the first input terminal for removing a
selected signal component from the first telecommunication signal and
thereby preventing audio reproduction of the selected signal component at
the Near-End;
a second input terminal for receiving a second telecommunication signal
from the Near-End;
a detection filter connected with the second input terminal and configured
for detecting the presence of a component of the second telecommunication
signal that corresponds with the component removed from the first
telecommunication signal; and
a decision stage connected with the detection filter and responsive thereto
for generating a control signal indicating the presence of audio signals
generated at the Near-End when the signal component is detected.
2. The apparatus of claim 1 wherein the removal filter comprises a filter
for removing a range of frequency components from the Far-End signal.
3. The apparatus of claim 2 wherein the selected range of frequencies is
above about 3,000 Hz.
4. The apparatus of claim 3 wherein the selected range of frequencies is
limited to a range between 3,750 Hz and 4,000 Hz.
5. The apparatus of claim 3 comprising an analog-to-digital converter
connected with said first input terminal and the removal filter comprises
an anti-aliasing filter in said analog-to-digital converter.
6. The apparatus of claim 2 wherein the detection filter comprises a
bandpass filter having a bandwidth that corresponds to the range of
frequency components.
7. The apparatus of claim 6 wherein the decision stage is responsive to a
predetermined threshold of signal activity detected by the bandpass
filter.
8. An echo canceler, comprising:
a first input terminal for receiving a first telecommunication signal from
a Far-End;
a removal filter connected with the first input terminal for removing a
selected signal component from the first telecommunication signal;
a second input terminal for receiving a second telecommunication signal
from the Near-End;
a detection filter connected with the second input terminal and configured
for detecting the presence of a signal component of the second
telecommunication signal that corresponds with the component removed from
the first telecommunication signal;
a decision stage connected with the detection filter and responsive thereto
for generating a control signal when the signal component is detected;
an adaptive filter connected to receive the first telecommunication signal
and configured for producing an estimated echo signal on the basis of
internal coefficient values, the adaptive filter having adaptation means
responsive to the control signal for adapting the internal coefficient
values;
summing means for removing the estimated echo signal from the second
telecommunication signal.
9. The apparatus of claim 8 wherein the removal filter comprises a filter
for removing a range of frequency components from the first
telecommunication signal.
10. The apparatus of claim 9 wherein the range of frequencies is above
about 3,000 Hz.
11. The apparatus of claim 10 wherein the range of frequencies is between
3,750 Hz and 4,000 Hz.
12. The apparatus of claim 8 wherein the detection filter comprises a
bandpass filter having a bandwidth that corresponds to the range of
frequency components.
13. The apparatus of claim 12 wherein the decision stage is responsive to a
predetermined threshold of signal activity detected by the bandpass
filter.
14. A teleconferencing terminal for providing full duplex communication
between a Near-End speaker at a Near-End and a Far-End speaker at a
Far-End, comprising:
a first input terminal for receiving a Far-End telecommunication signal
from the Far-End;
a removal filter connected with the first input terminal for removing a
signal component from the Far-End telecommunication signal, whereby a
filtered Far-End signal is provided;
a loudspeaker connected with the removal filter for providing the filtered
Far-End signal in audible form to the Near-End speaker;
a microphone for receiving an audible signal from the Near-End speaker,
whereby a Near-End signal is provided;
a detection filter connected with the microphone and configured for
detecting a signal component of the Near-End signal that corresponds with
the component removed from the Far-End signal;
a decision stage responsively connected with the detection filter for
generating a control signal when the signal component is detected;
an adaptive filter connected to receive the Far-End signal and configured
for producing an estimated echo signal on the basis of internal
coefficient values, the adaptive filter having adaptation means responsive
to the control signal for adapting the internal coefficient values;
summing means for removing the estimated echo signal from the Near-End
signal, whereby an echo-canceled signal is provided; and
an output terminal connected with the summing means for transmitting the
echo-canceled signal to the Far-End.
15. The apparatus of claim 14 wherein the removal filter comprises a filter
for removing a selected range of frequencies from the Far-End signal.
16. The apparatus of claim 15 wherein the selected range of frequencies is
above about 3,000 Hz.
17. The apparatus of claim 16 wherein the selected range of frequencies is
between 3,750 Hz and 4,000 Hz.
18. The apparatus of claim 14 comprising an analog-to-digital converter,
wherein the removal filter comprises an anti-aliasing filter in the
analog-to-digital converter.
19. The apparatus of claim 14 wherein the detection filter comprises a
bandpass filter for responding to signal activity within the selected
range of frequency components.
20. The apparatus of claim 19 wherein the decision stage is response to a
predetermined threshold of signal activity detected by the bandpass
filter.
Description
FIELD OF THE INVENTION
This invention relates to full duplex telecommunication systems. More
particularly, the present invention relates to a device that improves
double talk detection in a "hands-free" teleconferencing system.
BACKGROUND OF THE INVENTION
Electronic communication systems have become essential in the Information
Age. Teleconferencing, mobile communications, and Internet technology have
evolved from costly technological conveniences to necessary tools of
modern communication and commerce. One communication tool increasingly
exploited for its versatility and flexibility is the teleconferencing
terminal. Teleconferencing has enjoyed widespread application in both
personal and commercial communication contexts. The ability for several
individuals at a Far-End location to participate in group discussions with
several individuals at a Near-End location, has proved to be particularly
valuable.
A speakerphone, which includes a speaker and microphone sections, is often
integrated with a traditional telephone terminal to obviate the need for
the traditional handset. The speaker and microphone of the speakerphone
enable "hands free" operation, thereby making the telephone terminal
available for use by all parties within range of the speaker and
microphone. Due to the close proximity of the speaker and the microphone,
it is necessary to employ circuitry for preventing the microphone from
re-transmitting received audio signals produced by the speaker. Such
re-transmission would otherwise cause undesirable echoes or sustained
feedback oscillations. Some teleconferencing systems employ analog voice
switching or echo suppression circuits. These circuits disable, or
substantially attenuate, one of the respective send and receive
telecommunication channels in favor of the other. One result of that type
of echo suppression is undesirable clipping of voice signals during a
conversation. Additionally, background noises are often completely muted
during pauses in the conversation. Such muting can be undesirably
perceived as disconnection of the call.
Digital adaptive echo cancellation technology has been developed as a
favorable alternative to echo suppression. Adaptive echo cancellation
techniques require digital signal analysis. The Far-End analog signal is
converted to a digital signal, processed and then re-converted to an
analog signal for output to the speaker. The microphone signal is
similarly converted for digital processing. In the echo canceler, an
estimated echo signal is produced by a digital filter and then subtracted
from the microphone signal.
The filter used in the echo canceler to produce the estimated echo signal
is an Adaptive Finite Impulse Response (AFIR) digital filter. The AFIR
filter performs this function by convolving the received Far-End signal
with internal coefficient values. The internal coefficient values are
updated during a telephone conversation by an error correlation procedure,
such that when the estimated echo signal is combined with the Near-End
signal the echo signal is effectively canceled electronically. In order to
ensure that the internal coefficient values are accurately updated, it is
necessary to perform the error correlation procedure during periods of the
conversation when only the Far-End participant is speaking. If the
microphone signal contains a Near-End speech component, then the echo
canceler will adapt the internal coefficients of the AFIR filter to
attempt to cancel the Near-End speech as well as any reflected Far-End
speech.
Known echo cancelers incorporate "double talk" detection circuits to
identify conditions where both the Near-End and Far-End participants are
speaking and to suspend AFIR coefficient adaptation when such conditions
exist. The known detection circuits perform a comparison of the average
energy of the loudspeaker signal to the average energy of the microphone
signal. If the microphone signal level exceeds a predetermined proportion
of loudspeaker signal, then adaptation within the digital filter is
suspended. However, such double talk detection circuits are known to make
incorrect determinations due to sharp changes in echo path response,
changes in speaker volume, and the time varying properties of signals,
among other factors.
In view of the state of art as described above, a double talk detection
technique is desired which is capable of accurately distinguishing between
a Near-End speech signal transmission and an echo signal, in order to
effect echo cancellation or suppression more accurately and reliably.
SUMMARY OF THE INVENTION
A double talk detection system is provided for connection between a
Near-End and Far-End path of a telecommunication terminal. A first filter
is connected to the Far-End receive path for removing a selected component
from the Far-End signal prior to providing the Far-End signal to the audio
output device of the terminal. A second filter is connected with the audio
input device of the terminal for detecting a component of the Near-End
signal corresponding to the component removed from the Far-End signal. The
double talk detection system determines that Near-End speech is present in
the Near-End signal of the selected signal component is present in the
Near-End signal.
In a preferred embodiment, the first filter comprises an attenuating filter
for removing a range of frequencies from the Far-End signal. The second
filter comprises a bandpass filter and a detector for detecting the
presence of the range of frequencies within the Near-End signal. The
presence of the range of frequencies within the Near-End signal indicates
that Near-End speech is present, because such signal components are
removed from the Far-End signal and therefore, could only have been
generated at the Near-End.
The double talk detection system further comprises a control output
terminal for signalling that Near-End speech has been detected. The
control output terminal is connected to a control input of the echo
canceler in order to suspend adaptation of the internal coefficients of
the AFIR filter when a Near-End speech signal is detected.
Other aspects the present invention will become apparent to those skilled
in the art upon reading and understanding the following detailed
specification and attached drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a functional block diagram of a teleconferencing terminal
incorporating the double talk detection system of the present invention;
FIG. 2 is a graph showing the frequency response characteristic of a first
filter of the double talk detection system of FIG. 1, superimposed on the
frequency spectrum of a Far-End signal; and
FIG. 3 is a graph showing the frequency response of a second filter of the
double talk detection system of FIG. 1, superimposed on the frequency
spectrum of a Near-End signal.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT
Referring now to FIG. 1, there is shown a block diagram of a communication
terminal 10, such as a conference telephone. The communication terminal 10
is configured to provide two-way audio communication between a Near-End
speaker 8 and a Far-End speaker (not shown). A Far-End signal x(t) is
supplied to an input terminal 12 in analog form via a traditional
telephone subscriber line. An analog-to-digital converter, ADC 14, is
connected to receive the Far-End signal x(t) from the input terminal 12
and to convert the Far-End signal x(t) into a digital signal x(k). The
digital signal x(k) is provided by ADC 14 along signal line 16.
The signal x(k) is supplied by signal line 16 to an input terminal of a
removal filter 18. The removal filter 18 is configured to remove a
selected audio component from the digital Far-End signal x(k). In a
preferred embodiment, the removal filter 18 comprises a wideband lowpass
filter for substantially eliminating a range of frequency components of
x(k) above a selected frequency. Referring to FIG. 2, there is shown a
preferred frequency response characteristic 20 for the removal filter 18.
Also shown in FIG. 2 is a typical frequency spectrum 22 of a speech
signal, such as may be received by the communication terminal 10. As shown
in FIG. 2, the speech spectrum 22 includes a range of frequency
components, including components above 4,000 Hz.
The frequency response characteristic 20 of removal filter 18 is designed
to substantially attenuate, or reject, a range of frequency components of
the speech spectrum 22 above a selected cutoff frequency. For typical
speech signals, it has been found that removing components above about
3750 Hz, or in a range from 3750 Hz to 4000 Hz does not significantly
degrade the perceived quality of the speech.
In other embodiments, the filter characteristic 20 can be configured to
remove signal components within any selected range of frequencies that is
normally present in a speech signal. It is preferable to remove relatively
high frequency components because of the logarithmic relationship between
audio frequency and human pitch perception. At relatively high frequency
ranges, for example above about 3000 Hz, larger ranges of frequency
components can be removed from a speech signal without affecting the
perceived sound quality than at relatively lower ranges of frequencies.
Referring again to FIG. 1, the removal filter 18 provides a filtered signal
x(k)' upon signal line 20. A digital-to-analog converter 22 is connected
to signal line 20 to receive and convert digital signal x(k)' to an analog
signal x(t)'. The analog signal x(t)' is then provided to a loudspeaker 24
for audible transmission to the Near-End speaker 8.
Microphone 30 is located at the near end for receiving and converting the
Near-End speaker's audible speech 26 into an analog electronic signal
y(t). The microphone signal y(t) also includes an echo component because
of the echo path 28 between the loudspeaker 24 and the microphone 30. An
analog-to-digital converter 32 is connected to receive the microphone
signal y(t), to convert y(t) into a digital signal y(k), and to provide
the digital signal y(k) along signal line 34.
A summing junction 36 is connected to receive y(k) from signal line 34. The
summing junction 36 is further connected to receive an estimated echo
signal e(k) from an AFIR filter 38. At the summing junction 36, the
estimated echo signal e(k) is subtracted from the digital Near-End signal
y(k) to provide an echo-canceled signal y(k)' along signal line 40.
Further processing may be performed on the signal y(k)', such as residual
echo attenuation and comfort noise injection as are commonly performed in
connection with echo cancellation. A digital-to-analog converter, DAC 42,
is connected to signal line 40 for converting the echo-canceled signal
y(k)' into an analog signal y(t)', and for providing y(t)' at an output
terminal 44 for transmission to the Far-End, such as by the previously
mentioned telephone subscriber line.
The AFIR filter 38 generates the estimated echo signal e(k) by convolving
the received Far-End signal with an estimated echo path impulse response
determined by internal coefficient values. The AFIR filter 38 is connected
to signal line 20 to receive the Far-End signal x(k)' from which the
selected signal component has been removed by the removal filter 18. The
AFIR filter 38 is further connected with signal line 40 to receive the
echo-canceled signal y(k)'. The AFIR filter 38 executes a normalized
least-mean-square (NLMS) procedure to adapt the internal coefficients in a
manner that minimizes y(k)'. In order to obtain internal coefficients that
accurately model the actual echo signal 28, such adaptation procedure is
preferably performed when the Near-End speech signal 26 is substantially
absent from the microphone signal.
Detection filter 48 is connected to receive the digitized microphone signal
y(k). Detection filter 48 is further connected with a decision stage 54',
which is configured to produce an adaptation control signal on control
terminal 46 when the presence of a Near-End speech signal is detected
within the microphone signal y(k). The detection filter 48 comprises a
filter for responding to the presence, within the microphone signal y(k),
of a signal component corresponding to the signal component removed from
the Far-End signal by the removal filter 18.
Referring to FIG. 3, the detection filter (48) is preferably configured to
operate according to frequency response characteristic 52. The filter
frequency response characteristic 52 is selected to substantially
attenuate frequencies outside of a selected frequency range that coincides
with the range of frequencies attenuated by the removal filter 18. In the
present embodiment, the detection filter 48 includes a bandpass filter for
detecting the presence of signal components within a range of frequency
from about 3750 Hz to 4000 Hz. The detection filter 48 may further include
an averaging filter for producing an average of the signal level in the
passband over a time interval commensurate with the adaption interval of
the AFIR filter 38.
The detection filter 48 is connected to a decision stage 54 for receiving
the filtered microphone signal and for producing the adaptation control
signal in response to the filtered microphone signal having a non-zero
value, or a value above a selected sensitivity threshold. Because the
detection filter 48 is arranged to detect the presence of microphone
signal components corresponding the signal components that were removed
from the Far-End signal, then detection of such microphone signal
components is presumably due to audio activity originating at the near
end. In the present example as shown in FIG. 3, the presence of signal
activity within the range of 3750 Hz to 4000 Hz will cause the decision
stage 54 to issue the adaptation control signal at terminal 46. The AFIR
filter 38 will respond to the control signal by discontinuing to adapt the
internal coefficient values.
The terms and expressions which have been employed are used as terms of
description and not of limitation. There is no intention in the use of
such terms and expressions of excluding any equivalents of the features
shown and described or portions thereof. It is recognized, however, that
various modifications are possible within the scope of the invention as
claimed. For example, while removal filter 18 has been described as a
distinct component, it is noted that analog-to-digital converters, such as
ADC 14, commonly comprise low-pass filters for anti-aliasing. Hence,
removal filter 18 may be omitted as a distinct component setting an
appropriate cutoff frequency for the anti-aliasing filter of ADC 14.
Additionally, while the present technique has been described in connection
with a teleconferencing terminal, the principles of the invention are
generally applicable to any two-way telecommunication system wherein it
would be desirable to determine whether a signal traveling in one
direction comprises a signal component generated exclusively at one end of
the communication path. For example, the AFIR 38, removal filter 18,
detection filter 48, summing junction 36, and decision stage 54, which are
collectively designated as FIG. 1 as echo canceler 60, may desirably be
deployed in a digital telecommunication system for eliminating hybrid
echo. It should also be appreciated that various functional components of
the invention may be implemented as analog-electric circuits,
application-specific circuits, or preferably, as one or more
appropriately-programmed digital signal processing integrated circuits.
Top